tomevault-io Claude to Cursor Converter
Export Claude Code skills as Cursor-compatible rule files for sharing. All output goes to a dedicated repo — never modifies local Claude Code configuration.
Cursor Rule Format (.mdc)
Cursor uses .mdc files (Markdown with frontmatter) in .cursor/rules/:
---
description: One-line description (used for rule matching)
globs:
- "**/*.py" # optional: file patterns this rule applies to
alwaysApply: false # true = always loaded, false = matched by description
---
# Rule Title
Rule content here. Markdown format.
Key differences from Claude SKILL.md:
- Extension:
.mdc(not.md— Cursor's convention for rules) - No
name:field — filename is the identifier globs:for file-pattern matching (Claude uses description triggers)alwaysApply:replaces Claude's always-loaded behavior- No
references/subdirectory convention — inline or split into companion rules
Conversion Process
1. Classify the skill
| Type | Converts? | Notes |
|---|---|---|
| Persona (infosec-architect, cpa-tax-specialist) | Yes | Direct conversion, strong match |
| Workflow (tdd, write-a-prd, prd-to-plan) | Yes | Direct conversion |
| Code quality (code-consistency, code-quality) | Yes | Use globs for language-specific rules |
| Tool-specific (token-miser, budi-analytics) | No | Claude Code-only features |
| Infrastructure (fleet-manager, openclaw-ops) | Partial | Remove Claude-specific tool refs |
2. Transform content
- Strip YAML frontmatter, rewrite as Cursor frontmatter (description, globs, alwaysApply)
- Remove Claude-specific tool references (Agent tool, Bash tool, Read/Write/Edit, MCP)
- Inline small reference files (<200 lines) directly
- Large references → separate
.mdccompanion files - Replace "Claude Code" with generic "AI assistant" where contextual
- Add attribution:
<!-- Converted from Claude Code skill: <name> -->
3. Output structure
claude-to-cursor/
├── rules/ # Converted .mdc files
│ ├── tdd.mdc
│ ├── tdd-tests-ref.mdc # Large reference companion
│ ├── code-quality.mdc
│ └── ...
├── agents/ # Converted agent files
│ └── ...
└── scripts/
└── convert.sh # Batch conversion script
4. What to skip
- token-miser (Claude Code subagent routing — no Cursor equivalent)
- budi-analytics (Claude Code hook analytics — Cursor-specific)
- skill-updater (meta-skill for Claude Code skill management)
- inbound-sync (claude.ai ↔ Claude Code sync workflow)
- openclaw-ops (node_primary-specific deployment)
- fleet-manager (SSH/Ansible fleet ops — not IDE-relevant)
- ansible-hardening (infrastructure, not IDE-relevant)
Rules
- NEVER modify local Claude Code skills or configuration
- All output goes to the claude-to-cursor repo only
- Keep attribution on all converted files
- Test in Cursor before marking as ready
- .mdc extension for rules, .md for agents
Source: scoobydont-666/shared-claude-skills — distributed by TomeVault.